31 results about "Context model" patented technology

An image decoding method performed by a decoding device according to the present document comprises the steps of: deriving a context model for a sign flag of a current residual coefficient of a current block; and decoding the sign flag on the basis of the context model, wherein the context model for the sign flag is derived on the basis of a sign flag of a residual coefficient decoded before the current residual coefficient in the current block.

A coding method, a coder, a bitstream and a storage medium are provided. The decoding method comprises: parsing a bitstream, determining first information, the first information being used for indicating a position of a last non-zero coefficient of a current transform block; if the first information indicates that the last non-zero coefficient is located at a top-left corner of the current transform block, decoding a first parameter of the current transform block based on a first context model; wherein the first context model is different from a second context model, and the second context model is a context model used by a transform block decoded before the current transform block.

Some aspects of the present disclosure relate to systems and methods for performing automated tasks using a robotic system in a human-centric environment. According to a first aspect of the present disclosure, hierarchical object identification is used to generate a contextual model of the environment around the robotic system. According to a second aspect of the present disclosure, a semantic understanding of a task is determined in response to a user query. According to a third aspect of the present disclosure, a task-specific controller is used in combination with a general locomotion controller to execute task or sub-task specific processes in connection with completing a query. One or more aspects of the present disclosure may be used in combination with each other and / or may be used with additional systems and processes for performing automated tasks in a human-centric environment.

This application discloses a method, apparatus, and medium for video encoding and decoding, and a method for processing visual media data. A template region associated with a location is an aligned template region used for contextual deduction of a set of latent syntax elements for encoding the quantization transform coefficient level at that location. The set of latent syntax elements includes: a first latent syntax element for encoding a basic range level, a second latent syntax element for encoding a lower range level, and a third latent syntax element for encoding a higher range level. For a first location of a transform block, at least one context model is determined based on the first template region associated with the first location, for at least one syntax element. The at least one syntax element is decoded from the encoding information of the transform block based on the at least one context model. A first quantization transform coefficient level at the first location is determined based on the at least one syntax element.

This invention discloses a context-based adaptive entropy coding method for audio, comprising the following steps: First, the quantized codebook index sequence Q and quantization side information S of the audio signal are sequentially fed into a causal convolutional context model, a cross-band attention network, and a cascaded codebook dependency network for processing, to obtain temporal context feature vectors, frequency context feature vectors, and quantization-level context feature vectors for the audio signal in the time domain, respectively. Second, the temporal context feature vectors, frequency context feature vectors, and quantization-level feature vectors are input into a neural probability prediction model for processing, estimating the conditional probability distribution of each codebook index. Third, the predicted conditional probability distributions are input into an adaptive arithmetic encoder to generate the encoded bitstream, and necessary header information is added to the encoded bitstream to form a complete bitstream. This invention can further improve the compression efficiency of audio coding.

Coding a quantized transform block includes selecting a wavefront scan order for coding quantized transformed coefficients of the quantized transform block. The quantized transform block is of size N×N and the wavefront scan order is such that locations (x, y−1), (x, y−2), and (x, y−3) are sequentially coded and locations (x−1, y), (x−2, y), and (x−3, y) are sequentially coded, for at least one x and one y where 2≤x<N and 2≤y<N. A probability distribution is selected for coding a quantized transform coefficient of the quantized transform coefficients. A context model for selecting the probability distribution includes at least two immediate neighbors of the quantized transform coefficient in the wavefront scan order. The quantized transformed co-efficient is entropy coded using the probability distribution.

The present invention relates to an image encoding / decoding method and device. The image decoding method according to an embodiment of the present invention may comprise the steps of: binarizing a syntax element; determining a context model for the binarized syntax element; encoding the binarized syntax element on the basis of a probability value induced by the context model; determining whether to update the context model; and updating the probability value of the context model on the basis of whether to update the probability value.

A method of decoding video data includes determining, based on a color format of a picture of the video data, a context increment for a syntax element indicating a prefix of an x or y coordinate of a last significant transform coefficient of a color component of a block of the picture using one of a first context model and a second context model, and decoding a bin of the syntax element using context-adaptive binary arithmetic coding (CABAC) with a context determined based on the context increment for the syntax element.

The invention discloses a situation awareness type intelligent production scheduling decision-making method and system and a medium, and the method comprises the steps: mapping physical layer information in original data of a production environment into semantic layer features based on domain knowledge, and constructing a real-time production situation model containing physical state constraints and semantic decision-making features; according to the real-time production situation model and the optimization target, dynamically generating a guide instruction containing constraint conditions and task description; inputting the guide instruction into the trained core processing model to generate a plurality of preliminary scheduling schemes; sorting the preliminary scheduling schemes according to the weight of the optimization target in combination with a real-time production situation model, and selecting an optimal scheduling scheme; and executing the optimal scheduling scheme passing the automatic verification, continuously monitoring the actual execution condition, identifying deviation and feeding back deviation information to the core processing model for iterative optimization. According to the invention, real-time perception and autonomous decision making of a scheduling system on a production situation are realized, and a scheduling strategy is automatically adapted to multiple targets along with dynamic constraints.

A template region associated with a position is an aligned template region for context derivation of a set of potential syntax elements that are used to code a quantized transform coefficient level at the position. The set of potential syntax elements includes first syntax elements for coding base range levels, second syntax elements for coding low range levels, and third syntax elements for coding high range levels. For a first position in the transform block, one or more context models are determined respectively for one or more syntax elements, according to a first template region associated with the first position. The one or more syntax elements are decoded from coded information of the transform block according to the one or more context models. A first quantized transform coefficient level at the first position is determined based on the one or more syntax elements.

This invention discloses an SQL injection detection method and apparatus that integrates instantaneous behavioral features and long-term evolutionary features. The invention captures user requests and records metadata; associates user sessions through a multi-dimensional session identification strategy, constructs and updates a dynamic context model for each session in real time; extracts multi-dimensional feature vectors of requests, calculates static anomaly scores, and achieves high-risk instant attack determination; for non-high-risk requests, performs multi-scale analysis on the session temporal entropy spectrum sequence to obtain temporal anomaly scores; combines the dynamic context model to perform semantic consistency checks and correct the temporal anomaly scores; integrates the static anomaly scores, corrected temporal anomaly scores, and context factors to calculate a session risk score; determines session-level malicious attacks based on a dynamically adjusted adaptive session threshold, and executes a tiered response. This invention constructs a spatiotemporal collaborative detection framework of instantaneous static entropy analysis and session-level temporal entropy spectrum analysis, possessing both single-request deep semantic analysis and cross-request correlation detection capabilities.

The present application relates to the field of image coding and decoding. Disclosed is an image coding or decoding method, which can be applied to next-generation image or video coding and decoding standards such as JPEG-AI. The method comprises: acquiring a first feature of an image, wherein the shape of the first feature is \[C, h, w]; performing reordering on the first feature, so as to obtain a second feature, wherein the shape of the second feature is \[4C, h / 2, w / 2], the second feature comprises C sub-features, and the shape of each sub-feature is \[4, h / 2, w / 2]; performing recombining on the second feature, so as to obtain K third features, wherein the shape of a kth third feature is \[C, h / 2, w / 2], and the kth third feature comprises one channel of each sub-feature in the second feature, K being a positive integer greater than 2; and coding or decoding the image on the basis of the K third features. The present application can fully utilize adjacent pixel information in the same channel of a hidden variable and pixel information at the same position in neighboring channels, such that the hidden variable predicted by a context model is more accurate, thereby reducing the number of bits required for coding the hidden variable, and reducing the code rate.

The application relates to the technical field of personalized intelligent decision-making, in particular to a mobile terminal rich-situation perception and personalized intelligent agent decision-making method, which comprises the following steps: constructing a three-state monitoring mechanism based on a minimum perceptible energy threshold, wherein the three-state monitoring mechanism comprises a dormant state, a short-time monitoring state and a long-time monitoring state; in the short-time monitoring state, a light-weight Gaussian mixture model is used to perform coarse-granularity classification on a sampled audio frame, a binary label representing human voice or environmental sound is outputted, and a continuous audio stream is converted into a structured acoustic event; when the structured acoustic event is a non-silence event, a multi-sensor context is collected and inputted into a multi-modal context model to obtain a scene label; and the scene label, the structured acoustic event and an audio segment are jointly inputted into a personalized intelligent agent decision-making model to perform situation-level reasoning and generate a concise and executable situational guide. The method realizes a closed-loop processing of low-power continuous monitoring, deep situation understanding and personalized guidance.

The application provides a video decoding method and device, a video encoding method and device, and relates to the technical field of video coding and decoding. The video decoding method comprises the following steps: acquiring entropy coding data of a first syntax element of a first chroma block; the first syntax element is a syntax element for identifying whether a transform block of the chroma block is a non-zero block; acquiring spatial domain information of the first chroma block according to a neighboring decoded chroma block of the first chroma block; determining a target context model according to the spatial domain information of the first chroma block; and acquiring a value of the first syntax element of the first chroma block based on the target context model and the entropy coding data of the first syntax element of the first chroma block. Some embodiments of the application are used for improving the accuracy of a context model used for entropy coding of the value of tu_cb_coded_flag.

Some aspects of the present disclosure relate to systems and methods for performing automated tasks using a robotic system in a human-centric environment. According to a first aspect of the present disclosure, hierarchical object identification is used to generate a contextual model of the environment around the robotic system. According to a second aspect of the present disclosure, a semantic understanding of a task is determined in response to a user query. According to a third aspect of the present disclosure, a task-specific controller is used in combination with a general locomotion controller to execute task or sub-task specific processes in connection with completing a query. One or more aspects of the present disclosure may be used in combination with each other and / or may be used with additional systems and processes for performing automated tasks in a human-centric environment.

A real-time data analysis and context-aware anchor broadcasting method, device and medium, through the integration of multiple levels of technologies such as real-time game data acquisition, character behavior feature analysis, action event prediction, tactical intention inference, deep context model emotion calculation and natural language generation, not only can identify single character behavior, but also can understand the cooperative tactical relationship of multiple characters, and further combine the deep context model to generate time-continuous emotion indicators, so that the voice broadcast and expression rendering of the virtual e-sports anchor can be dynamically adjusted according to the game situation, improving the realism and immediacy of the broadcast, thereby achieving the technical effects of improving the ability to capture key tactical events, optimizing the continuity of virtual anchor emotion performance, and improving the immediacy of the broadcast.

A system and method for generating models from digital images in an interactive environment comprising a memory and a processor in communication with the memory. The processor captures or derives metadata for one or more digital images. The processor derives transforms from the metadata to align the digital images with one or more three-dimensional (“3D”) models of objects / structures represented in the digital image. The processor generates an interactive environment which allows a user to view a contextual model of each of the objects / structures in two dimensional (“2D”) and 3D views.

A method for determining thresholds for skipping entropy coding of latent features in image and video coding using neural networks is described. A threshold based on the mean of the estimated standard deviations of all latents is proposed. For autoregressive neural networks, it is proposed to have two separate entropy parameter estimation networks to avoid drift between context model parameter estimates computed during training and inference: an initial entropy estimation network used in entropy skipping of quantized latent coding and decoding, and a refined entropy estimation network used in arithmetic coding and decoding of quantized latents. Methods and systems for multi-stage entropy skipping are also proposed.

Some aspects of the present disclosure relate to systems and methods for performing automated tasks using a robotic system in a human-centric environment. According to a first aspect of the present disclosure, hierarchical object identification is used to generate a contextual model of the environment around the robotic system. According to a second aspect of the present disclosure, a semantic understanding of a task is determined in response to a user query. According to a third aspect of the present disclosure, a task-specific controller is used in combination with a general locomotion controller to execute task or sub-task specific processes in connection with completing a query. One or more aspects of the present disclosure may be used in combination with each other and / or may be used with additional systems and processes for performing automated tasks in a human-centric environment.

The present invention relates to an image encoding / decoding method and device. According to an embodiment of the present invention, the image decoding method may comprise the steps of: acquiring a bitstream including an encoded syntax element; determining the number of context models applicable to the encoded syntax element; determining one context model among the determined number of context models; and decoding the encoded syntax element on the basis of the context model, wherein the number of context models applicable to the encoded syntax element is determined to be less than the number of context models used in regular coding of the syntax element.

Methods and apparatuses for decoding a compressed image using a neural image compression network are provided. The method may include generating context parameters associated with a compressed image, the context parameters corresponding to a first area. The method may also include determining that a long range global dependency exists between a first latent and a second latent in a long-range global area within the compressed image and predicting a plurality of context features using a transformer-based long-range context prediction model. The method may then include reconstructing the compressed image based on the predicted plurality of context features.

The application relates to a video processing method. The method comprises the following steps: parsing a current coding bit stream from a video code stream; according to a current local index, calling a target context model corresponding to the current coding bit stream from a model group to which a current syntax element corresponding to the current coding bit stream belongs, to decode the current coding bit stream, to obtain current bit data corresponding to the current coding bit stream; the model group is constructed according to the correlation between different candidate syntax elements and the coding function modules corresponding to the different candidate syntax elements, each model group comprises at least one candidate syntax element and a candidate context model corresponding to each candidate bit data of the candidate syntax element; and in response to the completion of the coding of the coding bit stream in the video code stream, determining an original video according to the current bit data and bit data obtained by decoding the coding bit stream before the current coding bit stream. The method can improve the decoding efficiency.

This invention belongs to the field of image compression and computer vision technology, and discloses a cross-scale feature compression method for machine vision tasks. The method first extracts multi-scale features using a pre-trained feature pyramid network, then compresses them into two single-scale features through a hierarchical feature fusion network. Finally, a cross-scale entropy coding network combines a super-prior and a context model for joint probability modeling and entropy coding. A hierarchical feature reconstruction network then reconstructs the multi-scale feature pyramid. During the training phase, a rate-distortion joint optimization method is employed, simultaneously constraining the bit rate, feature reconstruction error, and detection / segmentation task loss, enabling the compression process to adaptively serve the machine vision task. Compared with existing image and feature compression methods, this invention provides directly usable features for downstream tasks without requiring image reconstruction or repeated feature extraction, significantly reducing bit rate and decoding computational complexity while maintaining or improving task accuracy.

The application discloses a mobile phone interaction system and method based on multi-modal biofeedback and dynamic context perception, a multi-modal bio-acquisition module integrates micro-expression recognition, skin electric response sensing and eye movement tracking sub-modules, and micro-expression, skin electric response and eye movement signals of a user are collected in real time; a dynamic context perception module constructs a dynamic context model by comprehensively considering environmental light, scene type and time information; a data processing and decision engine adopts a pre-trained machine learning model, multi-modal physiological characteristics and dynamic contexts are analyzed by fusion, emotional state, attention concentration degree and cognitive load indexes of the user are calculated, and personalized interaction instructions are generated; an interaction execution module adaptively adjusts mobile phone notifications, volume, display interfaces and health intervention functions; the application realizes an interaction paradigm change from passive instruction response to active state perception, significantly improves naturalness, intelligence and personalization level of the interaction, and has a positive digital health management auxiliary value.

The application discloses an Instant-NGP model compression method based on a context model, and the method comprises the following steps: constructing and training a context model customized for a hash representation of three-dimensional and two-dimensional mixing; estimating the parameter values in the hash table by using the context model; fusing the information of all index nodes generating hash conflicts in the estimated parameter values; and realizing entropy coding by using an arithmetic coding technology, so as to realize the purpose of model compression. The application introduces the context model to estimate the binary parameter values in the hash representation, so as to reduce the information entropy, and finally uses the arithmetic coding technology to realize the compression purpose.

Provided are a coefficient coding / decoding method, an encoder and a decoder. The method includes: parsing a bitstream by adaptive binary arithmetic coding using a first context model based on quantization coefficient positions to be decoded to obtain a value of a non-zero identification; if the value is a first value, parsing the bitstream by adaptive binary aritlunetic coding using a second context model to obtain values of X preset identifications; if the value of the X-th preset identification is the first value, determining a target order of exponential Golomb coding, and parsing the bitstream using an exponential Golomb decoding algorithm of the target order to obtain remaining values of reconstructed quantization coefficient absolute values; and determining reconstructed quantization coefficient absolute values corresponding to the quantization coefficient positions based on the value of non-zero identification, values of X preset identifications, and remaining values of reconstructed quantization coefficient absolute values.

Methods and apparatuses for decoding a compressed image using a neural image compression network may be provided. The method may include generating long-range context model parameters associated with a high resolution compressed image, the long-range context model parameters corresponding to a first area. The method may also include splitting the generated long-range context model parameters into a first number of context parameter blocks. The method may also include for each block in the first number of context parameter blocks, predicting respective context features using a long-range context model and respective context parameter blocks, wherein the long-range context model uses a corner-to-center latent decoding strategy or an edge-to-center latent decoding strategy to decode latents associated with the high resolution compressed image. Then, the high resolution compressed image may be reconstructed based on predicted context features.